OpenClaw’s developers announced OpenClaw 2.0 over the weekend, a major update that adds shared sessions, multi-user collaboration and tighter security controls for enterprise deployments.
Released as v2026.8.1, the update spans installation, messaging, memory, skills, models, automations, browser and native applications, plugins and security.
Peter Steinberger said on X on Aug. 31 that the team had spent two months trying to “build OpenClaw with OpenClaw” and had shifted from individual local coding harnesses toward team.openclaw.ai, a shared agent environment.
“Multiplayer coding + infinite compute with nodes and cloud sessions has been a game changer for how we build,” Steinberger wrote, adding that local harnesses now “feel like relics of the past.”
OpenClaw 2.0 rebuilds the browser interface around conversations as the main entry point, with files, approvals, settings and live agent activity accessible in the same workspace. The release also adds clearer visibility into tool calls, command activity, file diffs and long-running background tasks.
The redesigned settings area now includes agents, memory, plugins, MCP servers, devices, communication channels and device pairing. OpenClaw also consolidated model-provider administration, including credential status and, where available, quota, account balance, budget and spending information.
Shared cloud sessions let multiple users work inside the same agent context and history. Multi-user Gateways can track who created a conversation and which prompts were submitted by identified participants, while owners and administrators can control whether other users can read, suggest changes, work in draft mode or participate directly.
Colin Johnson, chief executive of marketing metrics software firm Solvely, said in an X post that the new multiplayer WebUI allowed two developers to open the same live session, view the same history and artifacts, and add context without recreating prior work. “We were working inside the same context,” he wrote.
OpenClaw 2.0 also introduces more granular security controls. Approvals can be tied to a specific request, command, session and person, command permissions can be restricted by arguments and working directories, and sessions can run under read-only, guarded, workspace or full-access modes, with the highest level limited to administrators.
OpenClaw said organizations can require sandboxed execution for sessions created by specific identities, and execution will fail if the required sandbox cannot be provisioned. The platform also supports Docker and Podman sandboxes, per-agent and per-session sandbox scopes, remote execution nodes and disposable cloud workers.
The project said a Gateway should still be treated as a single trust domain, and organizations needing stronger separation between business units or customers should run separate Gateway instances, or “cells.” Fleet tooling for managing those cells remains experimental.
OpenClaw’s documentation says sandboxing and execution approvals are off by default, and baseline configuration assumes a trusted single operator unless administrators impose stronger restrictions. The Secret Store also does not encrypt values at rest and relies on filesystem protections.
OpenClaw said 933 contributors, including 569 first-time contributors, participated in the release, which included more than 16,000 pull requests. The project said it is stewarded by the OpenClaw Foundation, an independent 501(c)(3), and should be understood as a foundation release rather than an OpenAI product.
Featured image credit
